XiDao Cookbook

Practical migration guides and production patterns for teams using an OpenAI-compatible AI API.

Why this repo exists

Most AI integration content stops at the first successful request. Production teams usually need help with harder problems:

• safe provider switching
• phased rollout and regression testing
• fallback and routing strategy
• cost control by workload
• observability when traffic spans multiple model options

This cookbook focuses on those higher-intent, production-facing problems first.

Sections

1. Migration and rollout

• switch from OpenAI API with minimal code changes
• verify which SDK assumptions your app depends on
• run rollout checklists before moving real production traffic
• separate output-quality risk from integration-risk during migration

2. Reliability and routing

• fallback versus blind retry behavior
• health-aware routing patterns
• latency-tier and task-tier model selection
• staged rollout patterns that reduce outage blast radius

3. Cost optimization

• prompt trimming
• model routing by task value
• caching repeated requests
• token and request visibility for cost analysis

4. Framework examples

• Python OpenAI SDK
• Node.js OpenAI SDK
• LangChain-ready notes
• Vercel AI SDK integration notes

Recommended first documents

• guides/migrate-from-openai.md
• guides/openai-compatible-rollout-checklist.md
• guides/reduce-api-costs.md
• guides/openrouter-alternative.md
• guides/llm-failover-routing-patterns.md
• guides/vercel-ai-sdk-openai-compatible.md

Repo relationship map

• xidao-python-examples → minimal Python usage examples
• xidao-nodejs-examples → minimal Node.js usage examples
• llm-failover-router-demo → code-first reliability and routing examples
• xidao-cookbook → migration, rollout, routing, and cost guides

Links

• Website: https://global.xidao.online/
• Support: [email protected]
• Telegram: https://t.me/ccyu085